Prompt Engineer
Writes, tests and improves prompts, output schemas and evaluation rubrics for LLM apps
- Repository
- Jeffallan/claude-skills
- Stars
- 12,176
- Licence
- MIT
- Files
- 7 files
What it does
Once installed, the agent approaches prompt writing like a methodical prompt engineer. It drafts prompts with clear sections for role, task, constraints and format, chooses patterns such as zero-shot, few-shot or chain-of-thought, and designs JSON output schemas and system prompts with guardrails. It also helps build test cases and scoring rubrics to measure whether a prompt actually works, and suggests changing one thing at a time when debugging. The skill is documentation-driven: its SKILL.md points to reference guides on patterns, optimization, evaluation, structured outputs, system prompts and context management that the agent loads as needed.
Great for
- Developers building an LLM-powered feature who want a systematic prompt design and testing process
- Data or ML staff writing evaluation rubrics and test suites for prompts
- Anyone refactoring an inconsistent prompt to get more reliable or token-efficient outputs
Skip it if
- Quick one-off questions where a simple prompt is enough, since the workflow adds process overhead
- Situations needing guaranteed model accuracy: the skill gives method and examples, not proof that a prompt will work on your data or on future model versions
Try asking
- “Refactor this summarisation prompt so it returns exactly three one-sentence bullets and test it on edge cases”
- “Design a JSON output schema and function-calling definition for extracting invoice fields”
- “Write a system prompt with a persona and guardrails for a customer support bot, including injection defence”
- “Create an evaluation rubric and 20-case test set to compare two versions of my classification prompt”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 7 original files, the MIT licence and an attribution note.
prompt-engineer.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip prompt-engineer.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip prompt-engineer.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Provide your current prompt, a few real example inputs and what a good output looks like, so the agent can build meaningful test cases
- Ask for one change at a time when iterating, as the skill itself recommends, so you can see what helped
- Check any model-specific advice against the provider's current documentation, since the skill's guidance may lag behind model versions
What's inside
- references/context-management.md
- references/evaluation-frameworks.md
- references/prompt-optimization.md
- references/prompt-patterns.md
- references/structured-outputs.md
- references/system-prompts.md
- SKILL.md
Redistributed unmodified from Jeffallan/claude-skills under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.